Coupled but Late: Turn-Taking Between Full-Duplex Speech Models in Unscripted Dialogue

2026-10-06

AI Summary

This study addresses the anomalous turn-taking timing observed when full-duplex speech models converse with each other, which stems from the absence of human-like cognitive buffering mechanisms. Building upon the PersonaPlex-7B model, the authors construct a script-free dialogue system operating under a shared clock via audio token exchange, and conduct comparative analyses using the Switchboard dataset alongside channel delay experiments. The findings reveal that although turn-taking between models exhibits coupling, the median response latency reaches 400–560 ms, substantially exceeding the 137 ms typical of human conversation. This demonstrates that current models rely on reactive waiting strategies rather than the predictive preemption capabilities characteristic of human interaction. These insights provide critical empirical evidence for optimizing full-duplex conversational systems.

Abstract

Full-duplex speech models are trained to converse with a person, but they are increasingly made to converse with each other, in self-play data generation, agent societies, and model-based evaluation. In that loop no human absorbs a timing error: each model's turn-taking is the other's input. We ask what timing the loop settles into. Two PersonaPlex-7B instances exchange audio tokens on a shared clock in unscripted conversation, and one floor-transfer rule is applied to them and to Switchboard. Their timing is coupled: re-pairing speakers across conversations destroys it. But the floor changes hands late, at a median of 400-560 ms against 137 ms for humans, and the last 120 ms of the partner's turn, where human projection places a tenth of its transfers, holds 1% of theirs. Delaying one direction of the channel shifts the response one-for-one and leaves the run-up to it empty, consistent with a reactive wait after the perceived end rather than the turn-end projection human timing requires.

Problem

• full-duplex speech models
• turn-taking
• unscripted dialogue
• self-play
• timing dynamics

Innovation

• Full-duplex speech models
• Turn-taking
• Self-play dialogue
• Reactive timing
• Floor-transfer